We plan to determine the incidence and prevalence of dementia among black and white community residents aged 65 and over in a five county (urban and rural) area in piedmont North Carolina, and to examine the extent to which race differences in health and sociodemographic characteristics explain differences in dementia rates. The sample will be drawn from the longitudinal Duke EPESE multistage probability sample (n=4,164). In the incidence study, all who have experienced a 2 point decline on a cognitive function screen over a three year period (c. 700) and all who have died in the interim (c. 600) will be evaluated. Participation in the probability study will be on the basis of stratified random sampling with replacement (n=342). Stratification will be based on cognitive functioning score. The selection procedure will maximize the opportunity of including persons with dementia. In both studies uniform diagnostic criteria for dementia based on DSM-III-R and for Alzheimer's disease (based on NINCDS-ADRDA criteria) will be developed using the standardized CERAD protocols. Weighted data will be used to project information to the five county area. Race, sex and age specific and adjusted incidence and prevalence rates will be determined. In addition, logistic regression will be used to examine the extent to which race differences in dementia rates are related to the presence of vascular and metabolic conditions and sociodemographic characteristics.